Diffusion-weighted MR imaging: Diagnostic accuracy in patients imaged within 6 hours of stroke symptom onset

PURPOSE: To evaluate the diagnostic accuracy of diffusion-weighted magnetic resonance (MR) imaging performed within 6 hours of the onset of stroke sym ptoms. MATERIALS AND METHODS: The authors reviewed the patient records and images from all patients hospitalized in a 10-month period in whom diffusion-weigh ted imaging was performed within 6 hours of the onset of strokelike symptom s (n = 22). Analyses included comparison of the initial interpretation of t he diffusion-weighted images with the final clinical diagnosis; blinded rev iews of computed tomographic (CT) scans and conventional and diffusion-weig hted images; and determination of lesion contrast-to-noise ratios (CNRs). RESULTS: Diffusion-weighted images indicated stroke in 14 patients, all of whom had a final diagnosis of acute stroke. Diffusion-weighted images were negative in eight patients, all of whom had a final clinical diagnosis othe r than stroke (100% sensitivity, 100% specificity, chi(2) = 23.00, P < .000 1). Blinded reviews yielded 100% sensitivity and 86% specificity for diffus ion-weighted MR imaging (chi(2) = 15.43, P < .0005); 18% sensitivity and 10 0% specificity for conventional MR imaging (chi(2) = 2.85, P > .2); and 45% sensitivity and 100% specificity for CT (chi(2) = 4.40, P > .10). Lesion p ercentage CNRs were 77% for diffusion-weighted imaging, 5.5% for CT, 9.8% f or T2-weighted MR imaging, and 3.1% for proton-density-weighted MR imaging (P < .002 for diffusion-weighted imaging vs others). CONCLUSION: Diffusion-weighted MR imaging is highly accurate for diagnosing stroke within 6 hours of symptom onset and is superior to CT and conventio nal MR imaging.